The letter x denotes mean values. Number of plans, 28 997 (2015) and 30 390 (2016). Number of nonopioid drugs, 30 (2015 and 2016). Denominator for the formulary restriction variables (percentage with prior authorization [PA], step therapy [ST], quantity limits [QL], and in the high-cost tier [HCT]) is the number of nonopioid drugs covered. The bottom edge of each box shows the first quartile, the top edge the third quartile. The line within the box shows the median. The bottom whisker is the lower adjacent value and the top whisker is the upper adjacent value. The circles are outlier values lying outside of the upper/lower adjacent values.
Data were not procured for 2017. HCT indicates high-cost tier; PA, prior authorization; QL, quantity limits; and ST, step therapy.
The letter x denotes mean values. The bottom edge of each box shows the first quartile, the top edge the third quartile. The line within the box shows the median. The bottom whisker is the lower adjacent value and the top whisker is the upper adjacent value. The circles are outlier values lying outside of the upper/lower adjacent values. NSAIDs indicates nonsteroidal anti-inflammatory drugs; SNRIs, serotonin and norepinephrine reuptake inhibitors; and TCAs, tricyclic antidepressants.
eTable 1. Description of Datasets Used in the Study
eTable 2. List of Drugs in Analytical Frame, by Drug Class
eTable 3. RXCUIs Indicated in the Beers List as “Potentially Inappropriate to Use in Older Adults”, and Excluded From Analysis
eTable 4. Cross-County Association of Non-Opioid and Opioid Formulary Exclusions With Opioid Prescribing, Using Alternate Definition of Opioid Prescribing Rate
Customize your JAMA Network experience by selecting one or more topics from the list below.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Rao T, Kiptanui Z, Dowell P, Triebwasser C, Alexander GC, Harris I. Association of Formulary Exclusions and Restrictions for Opioid Alternatives With Opioid Prescribing Among Medicare Beneficiaries. JAMA Netw Open. 2020;3(3):e200274. doi:10.1001/jamanetworkopen.2020.0274
Are Medicare Part D formulary exclusions and restrictions for opioid alternatives associated with increases in opioid prescribing?
Using county-level panel data throughout the United States, it was found that for each additional opioid alternative not covered in a county, the rate of opioid prescribing increased by 2.2% to 3.7% relative to the mean opioid prescribing rate. Formulary restrictions in the form of utilization management strategies and high-cost tier placements for opioid alternatives were not associated with increases in opioid prescribing.
The findings of this study suggest that lack of coverage of opioid alternatives may encourage higher rates of opioid prescribing.
Although there are many pharmacologic alternatives to opioids, it is unclear whether the structure of Medicare Part D formularies discourages use of the alternatives.
To quantify the coverage of opioid alternatives and prevalence of prior authorization, step therapy, quantity limits, and tier placement for these drugs, and test whether these formulary exclusions and restrictions are associated with increased opioid prescribing to older adults at the county level.
Design, Setting, and Participants
County fixed-effect models were estimated using a panel of counties across the 50 US states and the District of Columbia over calendar years 2015 and 2016. Data analysis was conducted from July 1 to September 23, 2019. The sample included 2721 counties in 2015 and 2671 counties in 2016 with sufficient data on Medicare Part D formulary design and opioid prescribing.
Main Outcomes and Measures
County-level opioid prescribing rate (number of opioid claims divided by the number of overall claims) and counts of excluded opioid alternatives and opioid alternatives with prior authorization, step therapy, quantity limits, and high-tier placements.
A total of 30 nonopioid analgesics were examined across 28 997 Medicare plans in 2015 and 30 390 plans in 2016. Medicare plans did not cover a mean of 7% of these drugs (interquartile range, 10%; lower to upper limit, 0%-23%). Among covered nonopioids, prior authorization and step therapy were uncommon, with fewer than 5% affected by prior authorization and 0% by step therapy. However, 13% of covered nonopioids had quantity limits (interquartile range, 10%; lower to upper limit, 0%-31%) and 22% were in high-cost tiers (interquartile range, 38%; lower to upper limit, 0%-50%). Increases in the number of nonopioids excluded on Medicare plans in a county were associated with increased opioid prescribing (effect size relative to mean, 2.2%-3.7%; P = .004). Conversely, increases in the number of opioids not covered on Medicare plans in a county was found to be associated with decreased opioid prescribing (effect size relative to mean, 0.8%-1.5%; P = .02). None of the utilization management strategies (prior authorization, step therapy, and quantity limits) examined or high-cost tier placements of nonopioids were associated with increased opioid prescribing.
Conclusions and Relevance
Lack of Medicare coverage for pharmacologic alternatives to opioids may be associated with increased opioid prescribing.
Between 1999 and 2011, opioid prescribing rates quadrupled in the United States.1 Despite modest declines since 2011, opioids remain widely oversupplied compared with preepidemic baselines in the United States and other regions of the world.2-4 Opioid overuse occurs among many populations, including older adults (age ≥65 years), most of whom are insured through Medicare Part D.5 Among the population of older adults, such use is associated with a number of potential adverse events beyond addiction and overdose, including falls, impaired motor coordination, and increased drug-drug interactions.6-8
Opioids are one of many treatments used to manage pain among older adults. Guidelines, including the 2016 Centers for Disease Control and Prevention guideline, encourage nonpharmacologic and nonopioid pharmacologic alternatives to opioids as first-line treatment for the primary care management of chronic pain.9,10 Nonpharmacologic alternatives to opioid use include therapies such as cognitive behavioral therapy, exercise therapy, and interventional treatments. Nonopioid pharmacologic alternatives include select drugs belonging to the following drug classes: nonsteroidal anti-inflammatory drugs (NSAIDs), antidepressants, anticonvulsants, topical analgesics, and muscle relaxants. Even though there is a broad armamentarium of nonpharmacologic therapies available for the treatment of pain and access to these therapies is relevant to understanding the use of pain drugs, in this study, our focus was limited to nonopioid pharmacologic alternatives and associated formulary policies.
Formulary policies, such as drug coverage, prior authorization, step therapy, quantity limits, and tier placement, are known to be important factors influencing drug use.11-13 Despite this knowledge, Medicare Part D prescriber’s guide on opioid overuse does not specify use of nonopioid pain relievers,14 and some have expressed concern that insurance companies may, at least inadvertently, contribute to opioid overuse through restrictions on access to safer alternatives.15,16
We quantified the extent of and variation in the formulary coverage and utilization management strategies for pharmacologic alternatives to opioids. We then evaluated whether these exclusions and restrictions were associated with increased opioid prescribing to older adults in the United States.
Data analysis for this study was performed from July 1 to September 23, 2019. The research ethics review board at IMPAQ International determined that this study did not qualify as human subjects research. The data included are publicly available and deidentified. Hence, institutional review board approval was not required. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies.
We used 2 main data sets to perform our analysis. First, we used the Centers for Medicare & Medicaid Services’ Prescription Drug Plan Formulary, Pharmacy Network, and Pricing Information Files (henceforth, Medicare formulary data) to analyze the prevalence of formulary exclusions and restrictions among opioid alternatives. We combined data from RxNorm, produced by the National Library of Medicine, with Medicare formulary data to obtain information on drug descriptions (eTable 1 in the Supplement).
Second, we used 2015 and 2016 Medicare Part D prescriber files (henceforth, Medicare prescriber data) to measure county level opioid prescribing rates. Since Medicare prescriber data are available at the zip-code level and Medicare formulary data are available at the county level, we merged the 2 using the Department of Housing and Urban Development’s zip code to county crosswalk. We then aggregated zip code–level prescriber data at the county level.
We used data from a variety of sources to construct covariates used in the regressions analysis, including US Census Bureau to derive county demographic information, Medicare Public Use Files to derive information on county Medicare fee-for-service population, the American Community Survey to derive information on county-level education and poverty rates, and the County Health Rankings & Roadmaps data sets to derive information on county level rates of primary care health care professionals. Details about each data set used in the study and the variables extracted from each source are provided in eTable 1 in the Supplement.
We used the Center for Disease Control and Prevention guidelines, published literature,17 and our own expertise to review the Part D formularies to identify opioid and nonopioid drugs used for the management of chronic pain (eTable 2 in the Supplement). Drugs were identified using RxNorm Concept Unique Identifiers. We additionally categorized nonopioid drugs into 5 groups: NSAIDs, antidepressants, anticonvulsants, topical analgesics, and muscle relaxants. Antidepressants included 2 subgroups: tricyclic antidepressants and selective serotonin reuptake inhibitors or serotonin and norepinephrine reuptake inhibitors. Any topical formulation of a drug that might otherwise be categorized as an NSAID, serotonin and norepinephrine reuptake inhibitor, anticonvulsant, or muscle relaxant was categorized as a topical analgesic.
For each nonopioid examined, we characterized its coverage and defined whether it was subject to prior authorization, step therapy, quantity limits, or placed in a high-cost tier, which we defined as tier 3 (preferred brand), tier 4 (nonpreferred drug), or tier 5 (specialty) of a typical formulary. Prior authorization is a requirement that prescribers obtain plans’ approval stating that it agrees to cover a particular drug. Step therapy is the requirement for a beneficiary to try a more cost-effective drug that has proven to work for most people, before trying a more expensive drug. Quantity limits are limits on the amount of a drug that a plan agrees to cover over a certain time period. Tier placements indicate the amount of copayment or coinsurance required for a drug. A large body of literature establishes that requirements, such as prior authorization, step therapy, and quantity limits, often reduce drug use,11-13 and higher cost-sharing requirements cause patients to substitute toward lower-value treatments.18,19
For each plan, a drug was classified as not covered/excluded if no RxNorm Concept Unique Identifier of a drug was listed on the formulary. Similarly, if any RxNorm Concept Unique Identifier of a drug was on the formulary with a given restriction (eg, prior authorization), the drug was classified as being covered with this restriction.
We analyzed the distribution (mean, median, and interquartile range [IQR], with lower limit values of the IQR defined as the 25th percentile minus 1.5 times the IQR and upper limit values defined as the 75th percentile plus 1.5 times the IQR) of the percentage of nonopioids with formulary restrictions across Medicare Part D plans, by restriction type and drug class. We analyzed data from 2015, 2016 (January formularies), and 2018 (March formulary). However, we were not able to include 2018 data in subsequent regression analysis owing to a lack of opioid prescribing data after 2016.
We excluded drugs from our analyses if they were identified in the 2019 American Geriatrics Society Beers Criteria as potentially inappropriate to use in older adults, regardless of any diagnosis or condition the person may have20 (eTable 3 in the Supplement provides a list of excluded RxNorm Concept Unique Identifiers of these drugs). The 2019 Beers Criteria recommendation on oral NSAIDs is to “avoid chronic use, unless other alternatives are not effective, and patient can take gastroprotective agent.”20[pg9] These oral NSAIDs were therefore included in the study.
We used ordinary least-squares linear regression models, with county-level fixed effects and additional time-varying covariates, to estimate the association between formulary exclusions and opioid prescribing, and formulary restrictions and opioid prescribing.
Following the convention of the Centers for Medicare & Medicaid, we constructed our primary outcome as county level opioid prescribing rate (ie, the number of opioid claims divided by the number of overall claims multiplied by 10021). As a sensitivity check, we also used an alternative definition of opioid prescribing, defined as the number of opioid claims divided by the number of Medicare enrollees multiplied by 100.
To construct our primary exposures for every plan, we first computed the number of nonopioid analgesics excluded; number of opioids excluded; number of drugs with prior authorization, step therapy, and quantity limits; and number of drugs placed in a high cost tier. To aggregate at the county level, we took a simple mean of these variables across plans in a county.
The use of county-level fixed effects obviates the need to control for county-level factors, such as rural or urban status, that are likely time-invariant over a 1-year time frame. Instead, to improve the precision of regression estimates and lower the probability of omitted variable confounding, we controlled for county-level demographic, socioeconomic, and health care–related covariates that may vary over time, including population, fee-for-service population, percentage of non-Hispanic white individuals, population between age 65 and 74 years, population between age 75 and 84 years, population aged 85 years or older, percentage of high school graduates, percentage of the population below the poverty line, and the rate of primary care health care professionals (number of primary care professionals divided by total population). Most of these variables have been established in the literature as significant factors to explain geographic variation in opioid prescribing.1 To test the sensitivity of our results to controlling for these covariates, we present both unadjusted and adjusted estimates.
The primary regression specifications have a sample size of 2721 counties (2015) and 2671 counties (2016), totaling 5392 county-year observations, in the unadjusted regressions. Counties in each year are fewer than total US counties (3007) on account of missing data in Medicare formulary and prescriber files. For 2015 data, 79 counties were missing from the formulary data file and an additional 206 counties could not be matched to the prescriber files. For 2016 data, 68 counties were missing from the formulary data file and the remaining 268 could not be matched to the prescriber files. In adjusted regressions, the slightly smaller sample size is on account on missing data in 1 or more covariates. All counties with missing data for variables included in a regression were excluded from analysis.
Analyses used 2-tailed, unpaired testing. A threshold P < .05 was used to establish statistical significance. Statistical analyses were conducted using Stata software, version 15 (StataCorp).
Thirty nonopioid drugs were examined across 28 997 plans in 2015 and 30 390 plans were examined in 2016. In Figure 1, we present the percentage of nonopioid drugs with formulary exclusions and restrictions across Medicare Part D plans, by pooling 2015 and 2016 data.
Nationwide, Medicare plans covered most nonopioid drugs examined; on average, plans did not cover 7% of nonopioids, although variation was present across plans (IQR, 10%; lower to upper limit, 0%-23%).
Medicare plans generally did not require prior authorization or step therapy restrictions for covered nonopioids. On average, 5% of covered nonopioids required prior authorization (IQR, 4%; lower to upper limit, 0%-12%) and 0% of covered nonopioids required step therapy (IQR 0%, lower to upper limit, 0%-0%).
The use of quantity limits and high-cost tiers was more prevalent. On average, 13% of covered nonopioids had quantity limits (IQR, 10%; lower to upper limit, 10%-31%). High-cost tier placement was most prevalent and also highly variable. On average, 22% of covered non-opioids were placed in a high-cost tier (IQR, 38%; lower to upper limit, 0%-50%).
In Figure 2, we examine the mean prevalence of formulary exclusions and restrictions across Medicare plans by year. The mean percentage of nonopioids not covered on plans’ formularies increased from 6.3% in 2015 and 2016 to 9.8% in 2018. Increases in restrictions were evident for all formulary restriction types, except for step therapy. In 2015, 26% of covered nonopioid drugs had at least 1 formulary restriction. By 2016, this restriction increased to 30%, and by 2018, to 36%.
In terms of formulary exclusions, the 2 drug classes with less than 100% mean coverage were topical analgesics and NSAIDs. On average, plans covered 89% of topical analgesics and 90% of NSAIDs. In Figure 3, we present the percentage of covered nonopioids with at least 1 formulary restriction by drug class. We pooled 2015 and 2016 data.
Formulary restrictions among covered nonopioids were relatively common among topical analgesics, anticonvulsants, and selective serotonin reuptake inhibitor or serotonin and norepinephrine reuptake inhibitor antidepressants. On average, 60% of topical analgesics were covered with at least 1 type of formulary restriction (IQR, 25%; upper to lower limit, 25%-100%). Analogously, 40% of anticonvulsants were covered with at least 1 type of formulary restriction (IQR, 0%; upper to lower limit, 33%-33%) and 39% of selective serotonin reuptake inhibitors or serotonin and norepinephrine reuptake inhibitors antidepressants were covered with at least 1 type of formulary restriction (IQR, 25%; upper to lower limit, 0%-75%).
In Table 1, we present the association between nonopioid formulary exclusions and opioid prescribing while controlling for opioid formulary exclusions. The number of nonopioids excluded from plans’ formularies in a county was significantly and positively associated with county-level opioid prescribing (unadjusted: β coefficient, 0.205; 95% CI, 0.127-0.284; P < .001; adjusted: β coefficient, 0.121; 95% CI, 0.040-0.201; P = .004). Thus, for each additional nonopioid excluded from plans in a county, the rate of opioid prescribing increased by 2.2% to 3.7% relative to the mean opioid prescribing rate, which was 5.5%.
The number of opioids excluded from plans’ formularies in a county was also significantly associated with county-level opioid prescribing. The number of opioids excluded from plans’ formularies in a county was significantly and negatively associated with county-level opioid prescribing (unadjusted: β coefficient, −0.085; 95% CI, −0.111 to −0.059; P < .001; adjusted: β coefficient, −0.044; 95% CI, −0.081 to −0.008; P = .02). In percentage terms, for each additional opioid excluded from plans in a county, the rate of opioid prescribing decreased by 0.8% to 1.5% relative to the mean opioid prescribing rate.
Using similar cross-county regressions, there was no statistically significant association between the number of covered nonopioids with step therapy or quantity limits or placed in a high-cost tier and county-level opioid prescribing, although increases in the number of nonopioids with prior authorization was associated with decreases in the rates of opioid prescribing (Table 2).
In eTable 4 in the Supplement, we reestimated Table 1 results with an alternative definition of opioid prescribing as a sensitivity check. As previously discussed, the opioid prescribing rate under the alternative definition was defined as the number of opioid claims as a fraction of total Medicare enrollees.
The results remained qualitatively similar. The number of nonopioids excluded from plans’ formularies in a county was significantly and positively associated with county-level opioid prescribing (unadjusted: β coefficient, 3.699; 95% CI, 1.616-5.782; P < .002; adjusted: β coefficient, 3.341; 95% CI, 1.177-5.504; P < .003). Thus, for each additional nonopioid excluded from plans in a county, the rate of opioid prescribing increased by 1.9% to 2.1% relative to the mean opioid prescribing rate, which was 179.97%.
Using this alternative definition, we also found that, for each additional opioid excluded from plans in a county, the rate of opioid prescribing decreased by 1.4% to 1.5% relative to the mean opioid prescribing rate. We also reestimated Table 2 results with this alternative definition of opioid prescribing and found that our findings remained unchanged; there was no statistically significant association between the number of covered nonopioids with step therapy or quantity limits or placed in a high-cost tier and county-level opioid prescribing, and increases in the number of nonopioids with prior authorization were associated with decreases in the rates of opioid prescribing.
Although the opioid epidemic is a complex phenomenon associated with several factors, one important factor in opioid prescribing is coverage and reimbursement policies by health insurers. This study makes 2 points to inform Medicare prescription drug policy. First, we highlight that drugs that can serve as effective and nonaddictive alternatives to opioids face substantial variation in coverage across Medicare drug plans. Second, using a county-level ecologic analysis of Medicare formulary and prescriber data, we found that, for each additional opioid alternative excluded by formularies in a county, the rate of opioid prescribing increased by 2.2% to 3.7% relative to the mean opioid prescribing rate. These results are notable because opioids are one of many pharmacologic treatments for pain, and the harder it is to access their alternatives within Medicare Part D plans, the more opioid use may ensue. These results suggest that policy makers can craft policies for Medicare drug plans to follow that emphasize increasing coverage or reducing restrictions on important opioid-alternative drugs.
Prior work, while descriptive, has characterized some of the variation in the coverage of nonaddictive alternatives with a focus on the treatment of lower back pain.17 There are several reasons why such variability may exist across plans, including a large and ever-evolving evidence base, the economic incentives faced by insurers and pharmacy benefits managers, and the administrative costs and complexities of managing formulary benefits for large and diverse populations.
The extent to which health care professionals may default to prescribing opioids when a particular nonopioid drug is excluded from coverage in a beneficiary’s plan depends also on the coverage status of other nonopioid alternatives that may be easily interchangeable with the excluded drug. For instance, alternatives within the same drug class may be more easily interchangeable. The interchangeability of products within a class varies across classes and there may be little substitutability in some classes, in which case prescribers may be more likely to default to opioids. In addition, prescribers are guided by many different considerations, including formulation, dosing, adverse event profile, cost, and tolerability, which may determine their prescription choices.
Our results build on other work examining Medicare coverage policy relevant to the opioid epidemic, including analyses of formulary restrictions for prescription opioids22 and opioid potentiators.23 Some studies have also looked at the association between formulary restrictions and use of alternatives to opioids11-13 and have generally found that restricting access to opioid alternatives reduces the use of the restricted drug and increases opioid use. Generally, these studies have tended to evaluate specific nonopioid drugs, often in the context of other payers. For instance, one study investigated prior authorization restrictions on pregabalin for the management of diabetic peripheral neuropathy or postherpetic neuralgia. Comparing 2 Medicaid state programs with prior authorization restrictions on pregabalin with 4 states with no restrictions, the investigators found a higher probability of pregabalin use, a lower probability of opioid use, and lower diabetic peripheral neuropathy or postherpetic neuralgia management costs in the unrestricted Medicaid state programs.11 However, in contrast to these prior studies, to our knowledge, ours is the first to characterize the coverage of a comprehensive set of nonopioid drugs across Medicare plans offered nationwide and analyze the implications of restrictiveness of access to these drugs at the county level for county-level opioid prescribing.
We did not find a statistically significant association between step therapy, quantity limits, or the use of high-tier placement for nonopioids and the likelihood of opioid use. However, we found that prior authorization for nonopioids was associated with decreased opioid use. One possible reason is that plans that apply prior authorization to nonopioids also use prior authorization to manage opioid use. However, model specifications that controlled for opioid prior authorization did not alter the findings. These findings could reflect biases on account of using ecologic variation in formulary restrictions and opioid prescribing to estimate the presented associations. For instance, the negative association between prior authorization for nonopioids and opioid use could reflect reverse causality bias; insurers may find it easier to increase prior authorization requirements on opioid alternatives in areas where opioid use is on the decline. Future work should analyze the mechanisms by which these formulary restrictions and opioid use are related.
Our analysis has limitations. First, the associations we presented are based on 2 years of publicly available data from 2015 and 2016, and the opioid epidemic and coverage policies have continued to evolve. The constraint with our study was the lack of access to more current prescriber data at the time of analysis. However, we had access to 2018 formulary data. We used these data to show that formulary restrictions for nonopioids apparently have been increasing among Medicare plans, making access to safe analgesics both a pertinent and timely policy issue.
Second, our ecologic analysis does not allow for causal inference, and it is possible that unmeasured or unmeasurable characteristics are associated with both formulary restrictions and opioid prescribing. Third, our analysis is limited to pharmacologic therapies, while coverage and reimbursement for nonpharmacologic treatments are also important to consider as part of comprehensive approaches to pain management.24,25 Fourth, these results are specific to the Medicare population and cannot be easily generalized to other populations among whom the potential for interchangeability across pharmacologic alternatives could differ.
A valuable extension of this work in the future would be to conduct a similar analysis at the plan level. Although a county, rather than plan, level measure of nonopioid formulary exclusions/restrictions may more accurately reflect an individuals’ access to nonopioids, averaging prescribing and restrictions at the county level reduces variation in the data.
Chronic pain affects millions of Americans and can be safely and effectively treated with many pharmaceutical and nonpharmacologic approaches. Although there are many barriers that inhibit unfettered access to these treatments, an important impediment for many individuals is insurance coverage and reimbursement policies that may discourage their use. In our analyses of county-level formulary and prescribing data for Medicare beneficiaries, formulary exclusions to nonopioids among Part D plans was associated with increased prescription opioid use. These findings underscore the importance of comprehensive coverage and reimbursement policies that target not only the oversupply of opioids but also the underuse of safer and more effective opioid alternatives.
Accepted for Publication: January 7, 2020.
Published: March 2, 2020. doi:10.1001/jamanetworkopen.2020.0274
Open Access: This is an open access article distributed under the terms of the CC-BY-NC-ND License. © 2020 Rao T et al. JAMA Network Open.
Corresponding Author: Tanvi Rao, PhD, IMPAQ, 1325 G St NW, Ste 900, Washington, DC 20005 (email@example.com).
Author Contributions: Dr Rao had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Rao, Kiptanui, Dowell, Alexander, Harris.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Rao, Kiptanui, Dowell.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Rao, Triebwasser.
Obtained funding: Rao, Kiptanui, Dowell, Harris.
Administrative, technical, or material support: Rao, Kiptanui, Dowell, Triebwasser, Harris.
Conflict of Interest Disclosures: Dr Rao reported receiving a small internal grant from her employer IMPAQ International, LLC, to help fund time and resources for this study. Dr Alexander’s contribution to this study was as a cofounder and principal of Monument Analytics, which was compensated by IMPAQ. This arrangement was reviewed and approved by the Johns Hopkins University in accordance with its conflict of interest policies. No other disclosures were reported.
Funding/Support: This study was funded by an internal grant from IMPAQ International LLC.
Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: Research assistance and feedback on early manuscript drafts were provided by Morteza Saharkhiz, PhD, Ping Chen, PhD, Tughluk Abdurazak, MS, and Bo Feng, PhD (IMPAQ International LLC). No compensation was received.
Create a personal account or sign in to: